On the equidistribution of some Hodge loci (Q2180849)

Given a nontrivial polarised variation \(\mathbb{V}\) of Hodge structure of weight \(2\) over a quasi-projective curve \(S\), with \(h^{2,0}=1\) (example: the cohomology of a family of \(K3\) surfaces) it is known that the Noether-Lefschetz locus, i.e.\ the set of points where the Picard number (the rank of the group of integral \((1,1)\)-classes) jumps up, is a countable dense subset of \(S\). This raises some further questions about the distribution of such points within \(S\). The main result is an asymptotic formula for the number \(N(\gamma,n)\) of points \(s\in S\) for which there exists an element \(x\in\mathbb{V}_{\mathbb{Z},s}\) of square \(-2n\) and class \(\gamma\in\mathbb{V}_{\mathbb{Z},s}^\vee/\mathbb{V}_{\mathbb{Z},s}\). Also, such points are equidistributed with respect to the measure given by integrating \(c_1(\mathcal{F}^2\mathcal{V})\). The formula applies in its precise form for variations where the local system \(\mathbb{V}_{\mathbb{Z},s}^\vee/\mathbb{V}_{\mathbb{Z},s}\) is trivial but the equidistribution result holds without this. \par There are interesting corollaries concerning the number of \(K3\) surfaces in a family admitting certain parabolic bundles or having elliptic fibrations of a particular type, and these questions were the starting point of this work. \par The approach is to consider the numbers \(N(\gamma,n)\) as intersection numbers between curves and Heegner divisors in an orthogonal modular variety. There are relations among the Heegner divisors (in a suitable compatification) arising from the existence of the Borcherds modular forms, which were determined by \textit{A. Petersen} [Modular forms on the moduli space of polarised \(K3\) surfaces. Amsterdam: Korteweg-de Vries Institute for Mathematics (PhD thesis) (2015)], and these yield the upper bound on \(N(\gamma,n)\). The lower bounds require an entirely different argument, relying on the results of [\textit{A. Eskin} and \textit{H. Oh}, Duke Math. J. 135, No. 3, 481--506 (2006; Zbl 1138.11011)]. Both parts are hard: in neither case can the earlier results be applied directly in a straightforward way, and the ideas needed for the proofs are in themselves very interesting. Nor are the corollaries for \(K3\) surfaces (and hyperkähler) manifolds immediate.